Artificial intelligence (AI) is changing the world of software development. According to Gartner® Magic Quadrant™ for AI Code Assistants (link resides outside of ibm.com), by 2028, 90% of enterprise software engineers will use AI code assistants, up from less than 14% in early 2024.* By using these solutions, developers can boost efficiency and overcome common challenges, such as managing complex legacy codebases, incorporating new programming languages and onboarding developers. 

In 2024, as part of the annual IBM watsonx™ Challenge, IBM invited its own employees to demonstrate how they could overcome these challenges using IBM watsonx Code Assistant™.

Thousands of IBM employees, spanning hundreds of project teams, experimented with the many watsonx Code Assistant capabilities, including code explanation, code documentation, code generation, and test case generation. The time savings that the teams achieved not only unlocked new possibilities in their work, but they also represented potential benefits for development teams everywhere.

* Gartner, Magic Quadrant for AI Code Assistants, August 19 2024. GARTNER is a registered trademark and service mark of Gartner, Inc. and/or its affiliates in the U.S. and internationally and is used herein with permission. Magic Quadrant is a registered trademark of Gartner, Inc. and/or its affiliates and is used herein with permission. All rights reserved.

One of the watsonx Challenge winning teams applied watsonx Code Assistant to an urgent work dilemma and achieved significant efficiency gains. This small team, a mix of veteran and recently hired product developers, inherited more than 50 undocumented code repositories containing about 750 JavaScript files. The team faced the prospect of weeks of manual analysis to understand and properly document the code base.

For the challenge, the team fed 9 files containing more than 1000 lines of code into the gen AI assistant and prompted the solution to produce one-sentence summaries of each file.

One of the team’s newer developers, Asher Scott, was already using watsonx Code Assistant to explain small pieces of code successfully, but wasn’t sure how it would perform with larger files. “We had no grand expectations, but we were pleasantly surprised by how good it was,” says Scott, adding that the solution was even effective at explaining advanced syntax written using shorthand coding techniques. “It helped me take my skills to the next level.”

When the team reviewed the summaries to check for accuracy, “We found it really accurate, and particularly accurate for documenting APIs,” says Scott.

Scott’s veteran teammate Greg Gasper also experimented with the code generation capability of watsonx Code Assistant. Gasper’s primary role is not programming, but he was able to enter a natural language prompt into the solution’s integrated AI chat and generate a script to identify files that needed documentation. “I turned it around in minutes,” says Gasper, “and that code exposed the 750 JavaScript files that we needed to document.”   

Based on the analysis of similar files, the team projects that the 9 files used in the challenge would take at least 3 minutes each, on average, to manually review and summarize. By using watsonx Code Assistant, they summarized each file in approximately 12 seconds*. That’s a time savings of more than 90%.

As the team applies watsonx Code Assistant to the rest of their code base, Scott explains that “we’re talking about work that would take weeks and getting it done in minutes.”

 

*Internal testing with optimal bandwidth and response time for watsonx Code Assistant.